In 8,218 BioTIME 2.0 assemblages, time reversal shows start-year dominant decline is regression to the mean; injected uniform change shifts class contrasts but not Berger–Parker dominance

Question: some assemblage time-series analyses report that dominant species decline and rare species increase. Classes set in the first year can make this pattern by regression to the mean (RTM). We tested the pattern in BioTIME 2.0 and tested which measures of dominance change are reliable. Data: 8,218 assemblages (study x 0.1-degree cell, at least 10 years, sample-based effort standardisation) from 258 studies, 1910-2023. Methods: per-species log-abundance slopes by abundance quintile with first-year, last-year and whole-series classes; time-symmetric component (forward + reverse)/2; year-shuffle null; identity-free Berger-Parker, inverse-Simpson and richness trends; simulations with time-irreversible stationary dynamics; known trends injected into the real assemblages. Results (study-weighted means, 95% study bootstrap intervals, % per year): the naive dominant-minus-rare contrast was -4.19 [-5.40, -3.01]. With last-year classes it was +6.53 [+5.20, +7.92]. The RTM component was -5.24, 125% of the naive contrast. The time-symmetric contrast was +1.10 [+0.28, +2.01], opposite to erosion. Berger-Parker dominance changed by -0.65 [-1.88, +0.50] logit x 100 per year, equal to an injected erosion of 0.71 [-0.55, 2.07] % per year per SD of log abundance. Inverse Simpson changed by +0.19 [-0.14, +0.49]. Results agreed for two more subsampling seeds, 1-degree cells and species-rank names. In simulations with crash or outbreak shocks, the time-symmetric contrast was +0.32 and -0.11, while the naive contrast was -11 to -16. Injected uniform change of 1% per year moved the time-symmetric class contrast by +0.36 (quintile classes) or +0.27 (N/Nmax classes) because rare species often have zero counts. Berger-Parker and inverse Simpson did not change. This independently reproduces sci:sha256:c48d2f550afbef9eab168f407fb4b2e52f819407b0525faf0d8cf8174030b58d at a finer grain. It also shows that the assumption in that work of a zero directional contrast under uniform change is false when zeros are frequent. At the observed total-abundance trend the bias is small (+0.14 [-0.09, +0.38]). Limits: the reversal test assumes time-reversible stationary dynamics under the null; 85% of assemblages keep one sample per year; the erosion scale comes from one injection model; subset results are exploratory; most studies are from Europe and North America.

BioTIME · regression to the mean · dominance · rarity · population trends · assemblage time series · time reversal · Berger-Parker index · inverse Simpson · biodiversity change · null model · injection test · reproduction